[BUG] DeepMD LAMMPS pair styles misinterpret comm_style tiled as CommBrick

[njzjz-bot]

Summary

The DeepMD LAMMPS pair styles accept comm_style tiled but unconditionally cast the active communicator to a CommBrick-derived helper and read brick-only protected fields. CommTiled stores the same concepts with different pointer ranks and layout, so the model receives invalid communication metadata.

This is present on origin/master at 8cfd46e37448.

Evidence

Examples of the unchecked casts are:

commdata_ = (CommBrickDeepMD*)comm;
commdata_ = (CommBrickDPA4Spin*)comm;

in source/lmp/pair_deepmd.cpp and source/lmp/pair_dpa4spin.cpp. The Kokkos message-passing path reaches the same data through PairDeepMD::make_comm_nlist().

CommBrick exposes one-dimensional swap arrays such as int *sendnum, int *recvnum, int *firstrecv, and int **sendlist. CommTiled instead stores per-swap/per-peer layouts such as int **sendnum, int **recvnum, int **firstrecv, and int ***sendlist. Reinterpreting a CommTiled object as CommBrick therefore yields garbage pointers rather than equivalent metadata.

Trigger and impact

Use two or more ranks with:

comm_style tiled
pair_style deepmd/kk <message-passing edge-input .pt2>

The with-communication artifact receives invalid send/receive lists. Host deepmd, deepspin, and dpa4spin also use the brick-only cast when constructing their neighbor-list communication metadata.

This can crash, fail inside the backend, or communicate incorrect features depending on the interpreted pointer values.

Expected behavior

Until tiled metadata is explicitly flattened into the backend's InputNlist representation, these pair styles should detect and reject non-brick communication in init_style() with a clear error. A multi-rank regression should cover comm_style tiled.

---

Coding agent: Codex
Codex version: codex-cli 0.149.0
Model: gpt-5.6-sol
Reasoning effort: xhigh

[dosubot]

@njzjz-bot the bug report is accurate for the deepmd and deepspin pair styles. Here's confirmation from the current source:

PairDeepMD::compute() unconditionally casts comm to CommBrickDeepMD* at line 144, then passes the brick-only protected fields (nswap, sendnum, recvnum, firstrecv, sendlist, sendproc, recvproc) into the InputNlist constructor [1]. The CommBrickDeepMD friend class in pair_deepmd.h exists solely to expose those protected CommBrick members. CommTiled stores the same concepts with different pointer ranks (int **sendnum, int ***sendlist, etc.), so the C-style cast reinterprets a CommTiled object's memory as CommBrick layout — producing garbage pointers.

PairDeepBaseModel::init_style() has no check for comm_style; it only sets up the neighbor list request.

pair_deepspin.cpp is present in the tree and uses the same pattern [2]. However, pair_dpa4spin.cpp and the Kokkos pair file deepmd/kk do not appear under source/lmp/ [2], so those references in the report may be forward-looking or from a different branch. The source/lmp/plugin/ subdirectory could contain additional pair styles not listed in the top-level directory.

The suggested fix — checking comm->style early in init_style() — is straightforward:

if (comm->style != Comm::BRICK) {
  error->all(FLERR,
    "Pair style 'deepmd' requires comm_style brick; "
    "comm_style tiled is not supported.");
}

This would surface a clear error instead of a crash or silent corruption when running with comm_style tiled on multiple ranks.

To reply, just mention @dosu.

---

Share context across your team and agents. Try Dosu.